Privacy-first by design

Your documents.
Your rules.
Zero compromise.

Most AI platforms demand a trade-off between intelligence and control. Smart Doc Assist eliminates that trade-off — delivering enterprise-grade document Q&A with auditable answers, provable accuracy, and data that never leaves your perimeter without your consent.

Request a Demo → See How It Works

Built for compliance-sensitive organisations

Smart Doc Assist
Secure
What is our data retention policy for client records in Australia?
Under the firm's Data Governance Policy, client records must be retained for a minimum of seven years from the date of last engagement, in accordance with regulatory requirements. Records classified as litigation-hold are exempt from routine deletion schedules.
data-governance-policy.pdf · pg 12
Who approved the latest amendment?
The latest amendment (v3.2, effective 14 March 2025) was approved by the Chief Compliance Officer with sign-off from the Managing Partner.
amendment-log-2025.xlsx · row 47
Ask about your documents...
Why generic GPTs fail in business

Off-the-shelf language models hallucinate confidently, cannot cite a page number, and require uploading sensitive documents to third-party servers. For legal, medical, financial, and compliance teams, this is not a limitation — it is a liability.

Three problems solved by design

Built different.
For a reason.

01

Accuracy & Reliability

Answers are grounded exclusively in your verified knowledge base — never the open internet. Every response can be trusted, traced, and audited back to its source document and page.

Verified knowledge base
02

Auditability

Every answer includes traceable citations with document name and page number. Admins can review which documents were used and the confidence score for each response.

Full audit trail
03

Data Privacy

No sensitive data ever leaves your secure environment. Only minimal, anonymised, and pre-screened excerpts are transmitted externally. Your raw documents stay where they belong — with you.

Sandboxed environment

How it works

Six steps.
Zero data exposure.

A purpose-built pipeline keeps sensitive data local at every step. The AI reasoning layer receives only the minimum anonymised context needed to generate an accurate, cited answer — nothing more.

Admins upload documents in 17+ formats — PDFs, Word, Excel, presentations, images, audio, and more. Each file enters your private, verified knowledge base under full admin control. Nothing is indexed without deliberate approval.

PDF DOCX XLSX → Knowledge Base
admin-controlled ingestion

Documents are automatically parsed, structured, and indexed entirely within your sandboxed installation. Headings, tables, and sections are preserved to maintain context. Nothing leaves your server during this stage.

Parse → Structure → Index
100% on-server · zero external calls

When a user asks a question, the system searches your verified knowledge base using multiple retrieval strategies simultaneously. Only the most relevant passages are selected — respecting tag-based access controls so users only receive answers from documents they are authorised to see.

Query → 847 docs → 9 results
access-controlled · multi-strategy retrieval

Before anything leaves your server, all personal names, organisations, locations, dates, and custom sensitive terms are automatically detected and replaced with anonymous placeholders. The AI reasoning service never sees the real identities in your documents.

"John Smith" → [PERSON_1]
names · dates · locations · custom terms

Only a small, anonymised text fragment is transmitted via an encrypted channel to the AI reasoning service. Never a full document. The reasoning service cannot reconstruct the original document or identify any person or organisation.

Anonymised 🔒 AI Reasoning
encrypted channel · fragments only

The AI response returns to your server, where the anonymous placeholders are replaced with the original values. Source citations — document name, page number, and section — are attached. The user receives a complete, verifiable, and accurate answer.

✓ Cited Answer → policy.pdf · pg 12
restored · cited · verifiable

Platform

A managed knowledge base.
Not just another chatbot.

SDA gives your organisation a verified, controlled knowledge base with admin oversight at every level.

Verified Knowledge Base Management

Admins build and curate a verified document library — not an uncontrolled web scrape. Upload documents in 10+ formats. Every file is processed, indexed, and made searchable under full admin control.

  • Admin-controlled document library
  • PDF, DOCX, PPTX, XLSX, and 10+ formats
  • In-browser document viewer with page-level citations
Documents
847
indexed
Formats
10+
supported
Status
Ready
verified
compliance-handbook.pdfReady
✓
amendment-log-2025.xlsxReady
✓
board-minutes-Q1.docxReady
✓
training-deck.pptxProcessing
65%

Organisation-Wide Access Control

Tag-based access control lets admins decide exactly which users can access which documents. Teams see only what they're authorised to — legal sees legal, HR sees HR. No tags assigned means no access, by default.

  • Tag-based document segmentation
  • "No tags = no access" security default
  • Multi-department, multi-team support
All Documents847 docs
—
Tag: Legal312 docs
filtered
Tag: Compliance189 docs
filtered
User's accessible docs94 docs
TBAC
Relevant results returned9 passages
secured

Intelligent Anonymisation

Sensitive information — names, dates, addresses, identifiers — is automatically detected and replaced before any data leaves the server. Admins can define custom anonymisation patterns for industry-specific terms their organisation needs protected.

  • Automatic PII detection and masking
  • Admin-defined custom patterns
  • Bidirectional — restored in final answers
IN
Original context
"John Smith signed on 14 March at 42 Pitt St, Sydney"
↓
PII
Sensitive entities detected
PERSON · DATE · LOCATION
↓
✓
Anonymised output
"[PERSON_1] signed on [DATE_1] at [LOCATION_1]"
↓
↩
Restored in answer
Original values re-inserted after reasoning

Full Admin Oversight

A dedicated admin dashboard gives your organisation complete visibility — total users, documents, questions asked, and resource usage. Review conversation history, export data, and monitor system health from a single pane of glass.

  • Usage analytics and activity charts
  • Conversation history review and CSV export
  • User management and role control
Users
7
6 active
Documents
66
all ready
Questions
824
all time
Daily Activity — Last 60 Days

No Sensitive Data Leaves

Document parsing, indexing, search, and PII detection all happen inside your sandboxed environment. The only data that ever leaves is a small, anonymised excerpt — stripped of all identifying information. Your raw documents never leave.

  • Sandboxed, isolated per-organisation environment
  • Raw documents never transmitted externally
  • Australian data residency (Sydney region)

Sandboxed

Each organisation gets an isolated, dedicated environment

Zero Raw Data Out

No sensitive information ever transmitted externally

Admin Control

Admins manage users, documents, access, and anonymisation rules

AU Hosted

Data residency in Sydney, Australia

0
Answer accuracy
Zero
Sensitive data transmitted
0
Document formats
0
Response time

Deployment

Your infrastructure.
Your choice.

Choose between hosted convenience or full on-premise data sovereignty.

Available Now
SaaS Subscription
Hosted Service
Up to 10 admins & 100 chat users · ready in hours
Contact Us
  • Deployed on DibaData secure servers
  • No infrastructure to manage
  • Updates and support included
  • Operational same day
  • Flexible billing with your own API key
Perpetual License
On-Premise
Up to 10 admins & 100 chat users
Coming Soon
  • Runs entirely on your servers
  • Full data sovereignty guaranteed
  • One day deployment engagement
  • Source-level customisation possible
  • Bring your own API key

Get Started

Ready to see SDA
on your data?

Book a 30-minute demo with real documents. No commitment, no generic slides — just your data, answered accurately.

Request a Demo → Contact Sales